Supervised Domain Adaptation for Extractive Question Answering in Spanish

ABSTRACT

Recent releases of pre-trained language models and Question Answering (QA) datasets have led to rapid improvements in Extractive QA. This paper describes the work done for QuALES, part of IberLEF 2022, a task to automatically find answers to questions in Spanish from news text related to Covid-19. We present an approach mainly centered on transfer learning applied to BETO and RoBERTa-base-bne based models. The models were fine tuned on different combinations of Spanish QA datasets. Our submission achieved third place in QuALES challenge for Exact Match and F1-Score metrics. © 2022 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).